用小波和分形理论处理短期负荷预测数据

Application of Wavelet and Fractal Theory on Data Treatment of Short-time Load Forecasting

针对电力系统负荷具有明显的分形自相似性特点,提出一种处理历史负荷中异常数据,并补充新数据的方法:采用小波理论分解信号,剔除异常数据;采用负荷之间的相似尺度的指标——分形维数来评估补充数据,得到适合该地区使用的选取阈值方法。实例表明,该方法对异常数据的处理及负荷数据的重构均有效,且具有一定可行性。

The paper presented a method of dealing with abnormal data in history load and supplementing new data aiming at the characteristic that there existed obvious fractal similarity in power system loads. Firstly, abnormal data were eliminated according to wavelet theory. Then,the supplement data were evaluated by fractal dimension which was the index of similar scale among loads; finally,the way of choosing threshold suitable to this area was got. Examples showed that this method had definite feasibility in dealing with abnormal data and reconstructing load data.